Hema A Murthy, Bhaskar Ramamurthi, Timothy A. Gonsalves,
K. Balamurugan, Shubha Augustine & Ch. Vijayalakshmi
TeNeT Group, Indian Institute of Technology, Madras 600036
E-mail: , .
Abstract
CygPlan is a GIS based access Network planning tool for TeNeT products. For the corDECT wireless local loop, CygPlan does base station propagation modelling taking into account obstacles. It also does near optimal placement of base stations in rural areas. The Model View Controller (MVC) architecture is used to design this tool. Following this design strategy different models in CygPlan have produced different views of the same.
1.Introduction
Planning and Designing an Access Network is a complex task. CygPlan is a Windows-based access network-planning tool that can be used to plan new installations. It examines the feasibility of an access network for a given area (for both voice and data) and can be used to generate a Bill of Quantities based on network traffic requirements.
The design of CygPlan is based on the MVC Paradigm (Fig. 1) [1]. The Graphical user interface (GUI) is the controller, which controls the models and views that are generated for a given access network.
In CygPlan's view of the network, the network consists of a set of Switching Centres (SC), which are in turn connected to Access Centres (ACs). Each AC can be made up of a heterogeneous collection of DIUs, BSDs, VRUs and IANs [2]. Facilities are provided in CygPlan to drag and drop appropriate Network Elements (NE) on to the map. Further subscriber characteristics can be supplied based on which CygPlan dimensions the plan both in terms of cost and equipment requirement. Finally CygPlan generates a Bill of Quantities which can further be modified by the user.
Additionally CygPlan also supports
a) Rural planning for corDECT Wireless Local Loop (WLL) and
b) Assists in determining the reach of Compact Base Stations (CBSes) in a given region for WLL.
In Rural planning, the tool determines the appropriate locations of Base Station Distributors (BSDs) under the constraint that maximum coverage is acquired.
Since corDECT is based on low power Line of Sight (LOS) communication, hills, tall buildings or trees can block or reduce the signal strength. CygPlan uses a prediction model for signal strengths at different distances from the CBSes.
2.Design
2.1.Overall Design
The MVC user interface Paradigm splits applications or interfaces into three parts: the model, the view and the controller (as shown in Fig.1). User interacts with the controller and makes changes to the model, which is then reflected in the view.
Figure 1. MVC Architecture
The Controller has reference to both the View and the Model and is the base of the triad. The model operates on the underlying data, the view is a representation of the operation of the model on the data, and the controller is the set of routines, which maps between the two. This framework offers a very flexible scheme for GUI development.
In CygPlan, we have a single GUI which directly interacts with different models which in turn generate a number of different views (Fig. 2) of the Access Network that is planned. The heart of the GUI consists of a Network Element Data Structure (NEDS) which is manipulated by different models.
Data flow may differ depending on what kind of action the planner performs. The planner can apply different kinds of filters on the data to generate different views of the network. Using the NEDS as an interface the GUI interacts with the Database and Excel.
Figure 2: MVC Architecture of CygPlan
The changes brought about here are in turn reflected in different views of CygPlan. The Bill of Quantities, Plan Summary, detailed reports for every network elements, the Propagation Model and the Rural Planning are different types of views generated by CygPlan.
In the next few sections we discuss the different models and views that are supported in CygPlan.
2.2.GIS
Dimensioning of CygPlan is done in the Geographic Information System (GIS) model. The design of GIS in CygPlan is based on the raster model. The map on paper is scanned and stored in the computer in the form of a bitmap.
The map is divided into a collection of grids or cells. The planner has to input the GIS information at the origin of the map. Once the GIS information for the origin and scale is defined the user can obtain information about any point on the map, either in lat-long or kilometers. The current version of CygPlan only allows planar maps, so the interpolation is primarily linear in scale.
Distance computations are also done in CygPlan. This is done using the Euclidean metric. Further, this view also supports dynamic change of scale on the plan.
2.3 Reports and Bill of Quantities Generation
One of the models in this planning tool comes up with a detailed configuration of the system to be deployed, a cost and bandwidth estimate and finally a bill of materials for the same (Fig. 3). The planner inputs a set of parameters to the system.
Figure 3: Bill of Quantities
The planner input includes:
· The number of subscribers for each system/subsystem and traffic characteristics of subscribers
· The DIUs, IANs, VRUs [2] that make up each Access centre.
· The subscribers for each IAN, BDSU and HSDU [2].
· The type of backhaul-radio, fiber, twisted pair or leased lines [2].
· Terrain objects like buildings and skylines.
· Generalization Positioning System (GPS) data for location of Network elements.
CygPlan has default cost values (re- configurable constants) and hardware constraints for all components. These can be overridden dynamically by the network planner.
This view checks the hardware and bandwidth constraints and provides network deployment cost and bandwidth estimates for the resultant plan, taking every detail into consideration including power supply requirements, cabinet sizes for devices, backhaul budgeting, towers and other infrastructures and miscellaneous details required to set up an access network. The computations are done at the back end by MS Excel and based on these computations the reports are generated by this view. Fig. 4 shows a summary of the plan generated.
Figure 4. Reports (SC report)
These reports form one of the views generated by CygPlan. The reports can further be filtered and viewed by the planner in the required format. The Bill of Quantities is another view that the tool generates using this model.
The Bill of Quantities provides detailed information about the quantities of the items to be purchased and their costs. A key feature is that the planner can change any of these and the bill will be automatically revised.
2.4.Optimum Placement
2.4.1.Optimum Placement of CBSes and RBSes
Figure 5: Potential Exchange locations (before finding the optimum placement)
For wireless installations in rural areas, potential exchange locations are available (which are the preferred positions for network nodes) in SDCAs (Fig.5). The planner would like to know the optimum locations to place the BSDs such that the signal covers maximum possible area, with a limited (fixed) number of BSDs. Manually dragging and dropping the BSDs for the coverage each time cannot be a practical solution for this problem.
CygPlan uses a modified k-means-clustering algorithm to determine the optimum placement. The algorithm for clustering is given in Fig. 6.
Algorithm Clustering
1. Get the list of potential locations
2. Get the required number of BSDs that
can be supported by DIU, say MAXn
3. Get the radius of the circular range
covered by the BSD
4. Divide the map into cells of specified
size
5. Randomly choose MAXn unique BSD
locations
6. Find the BSD nearest to each cell.
7. Cluster the cell w.r.t. the list chosen in
step 5
8. Find the centroid of each cluster
9. Find the nearest available BSD location
for each centroid.
10.Repeat the step 6, 7, 8 until all cells are
covered by the BSDs or percentage of
coverage does not increase significantly
from iteration to iteration.
11.Check the coverage obtained. If
required coverage were not attained do
step 2 through 9 by replacing BSDs by
RBSes in steps through 9 with MAXn
ranging from 1, 2, 3, ...n. n is determined
by the coverage required.
Figure 6: Clustering Algorithm
Further the planner has the provision
of using RBSes to cover small areas that could not possibly be covered using the available or permitted number of BSDs.
Fig. 7 shows the optimal locations for BSDs such that the area chosen for planning is maximally covered. The view generated by this model is shown in Fig.7.
Figure 7: Selected BSD, RBS locations with coverage
2.4.2 Plan Generation
The selected network elements are represented as a Directed Acyclic Graph (DAG) [3] rooted in the Head Quarters (Fig. 8). The Head Quarters forms the SC. SC holds an AC and a DIU. In addition to this, the AC can support 4 other slave ACs [2].
Figure 8: The selected network elements represented as a DAG
After finalizing the plan, the planner can generate a plan. All the details like subscriber information, backhaul, tower type, ISP interconnect, PSTN Interconnect, server, router, etc can be entered by the planner as required.
Further the tool is flexible in that, at any point in time, the automatic dimensioning can be overruled by manual dimensioning of the network.
2.5.CBS Propagation
2.5.1 Coverage Prediction
A coverage prediction tool is necessary for efficient planning of any wireless network. To predict the signal strength accurately in a microcellular network, information about the locations of buildings and their dimensions is required. However, this information may not always be available and would involve considerable effort to acquire. Further, the costs involved are also high. On the other hand, it is easier to get an estimate of the average height of the buildings (skyline) in various regions around a cell site. This data is sufficient to come up with a reasonable estimate of the signal strength.
CBS Propagation is another Model for processing the data in CygPlan. In this, the coverage is calculated and displayed on the basis of the signal strength at different places rather than the distance from the exchange locations.
The planner interacts with the GUI to input the data corresponding to the known obstacles in the path of the signal. The propagation prediction model provided in CygPlan allows the user to load a map and fill the areas around the Base Station sites with skyline information. The map is divided into square cells, the size of the square being specified by the user. The signal strength at each of these grid cells is calculated using a propagation model based on ray tracing (discussed in the following section). The grid cells are colored according to the best signal strength received in each square from all the Base Stations.
Figure 9: Coverage obtained using the Propagation Model
The view generated by this algorithm is a colour-coded map based on signal strength. The signal strength decreases from the centre of the cell to the periphery (Fig. 9). Observe that the signal strength changes significantly behind a building.
2.5.2. The Propagation Path Loss Model
The received signal strength is a function of the propagation path loss (which in turn is a function of factors like Base Station tower height, type of built-up area (urban, suburban, rural etc.,)) and antenna gains of the transmitter and receiver. The antenna gains, Base Station tower height (from ground), type of built-up area and the height at which the received signal strength is to be measured, can be specified by the user.
The path loss model has been developed using ray tracing. The signal transmitted by the Base Station travels partly through free space (LOS distance) and partly through built-up space (NLOS distance) before reaching the wallset antenna. In many models in the literature, a loss exponent greater than 2 is used for the NLOS part (e.g., an exponent of 3.5 is suggested in [4]). Such a model implies that doubling in length of the NLOS part, say from 1 km to 2 km, and will imply an additional loss of only 10.5 dB. This is contrary to experience, where the loss in the NLOS part is seen to increase linearly (in dB) with the distance. This is due to the fact that more walls are crossed as the distance increases and each wall leads to a fixed loss (on the average). To account for the higher loss seen in practice due to the NLOS path, we introduce a term ad2 in addition to the free space path loss, where d2 is the NLOS distance (Fig. 10). The expression for path loss at 1.9GHz then takes the following form:
Figure 10: Signal path from BS to WS
Path Loss = 38 + 20logd +a d2
Where d is the total distance (d = d1+ d2). The NLOS distance d2 is a function of the tower height t and skyline s, and the height of subscriber antenna.
The factor a reflects the density of the built-up space. The value of a is fitted empirically to the measured signal strength in the area chosen for planning. In general, a increases with urban density
3. Conclusion
In this paper, a Network planning tool for TeNeT products which uses the MVC paradigm was used. The advantage of using the MVC paradigm enables seamless integration of new strategies and new technologies for planning.
The software is developed in such a way that the planner can customize the data and parameters without changing the existing design. Additional features can be added with out affecting the existing modules. The use of MVC architecture enables us to add more views and models to CygPlan without disturbing the existing views, models or the GUI.